Transcripts for a whole YouTube playlist or channel
To get the transcripts of a whole YouTube playlist or channel, open vidwords.com/bulk, paste the playlist link or the channel's @handle, press Find videos, then Extract all transcripts. Up to 500 videos are listed for free with a free account (25 when signed out), with the exact cost shown before you start. Every transcript downloads as one TXT, CSV or JSON file, or as a ZIP of SRT or VTT files. Running the extraction needs a Starter plan or above, at one Cloud Request per video.
Either one opens the bulk extractor with the right tab already filled in. Press Find videos there. Nothing is spent until you extract.
The 30-second version
- Open Bulk extract and pick the Playlist or Channel tab, or use the boxes above.
- Paste any link containing
list=…, or an@handle, a channel URL or aUC…id. Press Find videos. Listing is free and covers up to 500 videos once you are signed in (25 signed out). - Choose the caption language and press Extract all transcripts. Progress shows per video, and a video that fails is listed with its reason and never charged.
- Download the whole run as one file (
.txt,.csvor.json) or as one file per video in a.zip, which is the only option for.srtand.vtt. - Automating it?
POST /api/channels(Starter and up) turns a handle into video ids; feed those toPOST /api/transcriptsin batches of 50.
Channel or playlist — which do you want?
Both run through the same bulk extractor and cost the same, so the only question is how the videos are grouped:
- A channel gives you a creator's most recent uploads — the right choice for competitor research, or for building a knowledge base from everything one person has published.
- A playlist gives you a curated, ordered set — the right choice for a lecture course, a conference track, or a podcast archive, where sequence matters and you don't want the creator's unrelated uploads.
A video that belongs to a playlist can be reached either way; pasting a URL that contains list=… resolves the playlist rather than the single video.
Extracting a channel in the browser
- Open Bulk extract and pick the Channel tab.
- Paste the channel's
@handle(e.g.@mkbhd), its URL, or aUC…channel ID — VidWords resolves the channel ID for you. - Hit Find videos. The latest videos (up to 500) are listed for free in the Bulk interface; successful transcript processing is metered separately.
- Click Extract all transcripts (Starter and above). Progress shows per video; failures (e.g. caption-less videos) are listed per video and never charged.
- Download everything as one combined
.txt,.jsonor.csv, or as a.zipwith one file per video.
Extracting a playlist in the browser
- Copy the playlist link — anything containing
list=…works, including a video URL that is part of a playlist. - Open Bulk extract → Playlist tab → paste → Find videos (free).
- Review the resolved list — title, channel, and count, with a note showing exactly how many Cloud Requests extraction will use.
- Hit Extract all transcripts, then download the combined file or the ZIP.
What the download looks like
| Format | Structure | Best for |
|---|---|---|
.txt | Each video under a heading with its title and youtu.be link, then its captions as timestamped lines | Reading end to end, searching, or pasting into an LLM |
.json | A videos list with per-video segments, timings, title, author and language, plus a failed list | Scripts, pipelines, and RAG indexes |
.csv | One row per caption line: video_id,title,start,duration,text | Spreadsheets and timestamp filtering |
.zip | One numbered file per video in the format you pick, .srt and .vtt included | Subtitle files for a whole series |
An illustrative slice of the combined .txt (made-up title and words):
==== Lecture 1: Introduction (https://youtu.be/VIDEO_ID) ====
[00:00:00] welcome to the course, today we'll cover
[00:00:04] what you need before lecture two
What people build with this
- Course notes — turn a 40-lecture playlist into one document you can search across.
- Podcast quote mining — search every episode for a phrase, then jump to the exact timestamp.
- Competitor and creator research — read what a channel actually says, at volume, instead of skimming titles.
- LLM knowledge bases — the combined JSON drops straight into a retrieval pipeline.
Over the API
The channels endpoint (Starter plan and higher) turns handles into video lists, which you then feed to the transcripts endpoint in batches of 50:
curl -X POST https://vidwords.com/api/channels \
-H "Authorization: Basic YOUR_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{"ids": ["@mkbhd"]}'
The response includes channelId, the channel title, and each video's id — pipe those ids into POST /api/transcripts.
Limits and practical notes
- Channels and playlists both resolve up to 500 videos per request, newest first for a channel and in playlist order for a playlist.
- One successful transcript uses one Cloud Request. Resolving the video list in Bulk is free, so you can check the size of a job before committing to it.
- Running the extraction needs a Starter plan or above. On Free you can still list a playlist and open any single video from the list, at one of your 25 monthly Cloud Requests each.
- Private and members-only videos can't be read, and some videos legitimately have no transcript (captions disabled, music-only, age-restricted). These come back as per-video errors rather than failing the whole batch, and they are not charged.
- Past 500 videos the listing is capped rather than paged — over the API the response comes back with
truncated: trueso you can tell it happened. For a back catalogue larger than that, split the work into narrower playlists.